Motor nerve terminals of arthropods provide excellent models for study of synaptic transmission, and their ultrastructure can be investigated in the same endings from which physiological recordings have been obtained. An experimental procedure for marking a recording site for subsequent ultrastructural analysis is described. The most commonly used procedure for ultrastructural analysis has been serial sectioning and three-dimensional reconstruction. This procedure has the advantage of providing information about the entire nerve terminal, including quantitative information on number, sizes, and relative positions of individual synapses and presynaptic 'active zones'. However, several errors may be generated in the process of viewing the sections and making the reconstruction; these errors can in principle lead to overestimation of synapse and active zone size. The errors become relatively more serious for smaller structures. Procedures for alleviating some of the possible errors are outlined. It is desirable to have additional information from other methods, such as freeze-fracture replication, to guide analysis of reconstructions from serial sections. Combined physiological and ultrastructural analysis of arthropod terminals has shown that each terminal has many small synapses, differing in size and in number of active zones, and that in some terminals, many of the observed synapses have a very low probability of transmission when nerve impulses occur at low frequencies.